COSMIC Database Toolkit

A scientific computing skill for querying COSMIC (Catalogue of Somatic Mutations in Cancer) — the world's largest database of somatic mutations in human cancer, maintained by the Wellcome Sanger Institute. COSMIC Database Toolkit helps you search for cancer-associated mutations, retrieve mutation frequency data, and analyze mutational signatures across cancer types.

When to Use This Skill

Choose COSMIC Database Toolkit when:

• Searching for somatic mutations associated with specific cancer types
• Looking up mutation frequencies for known cancer genes (TP53, KRAS, EGFR)
• Retrieving cancer gene census data for oncogenes and tumor suppressors
• Analyzing mutational patterns and signatures in specific tumor types

Consider alternatives when:

• You need germline variant data (use ClinVar or gnomAD)
• You need clinical trial data for cancer drugs (use ClinicalTrials.gov)
• You need gene expression in cancer (use TCGA via GDC)
• You need drug sensitivity data (use GDSC or DepMap)

Quick Start

Copy
claude "Find the most common TP53 mutations in lung cancer from COSMIC"

Copy
import requests

# COSMIC API requires authentication
# Register at cancer.sanger.ac.uk for API access
headers = {
    "Authorization": "Bearer YOUR_API_TOKEN"
}

# Search for TP53 mutations in lung cancer
url = "https://cancer.sanger.ac.uk/cosmic/api/v1/mutations"
params = {
    "gene": "TP53",
    "tumour_site": "lung",
    "sort": "count:desc",
    "limit": 10
}

response = requests.get(url, headers=headers, params=params)
mutations = response.json()

for mut in mutations:
    print(f"Mutation: {mut['mutation_cds']}")
    print(f"  Protein: {mut['mutation_aa']}")
    print(f"  Count: {mut['count']} samples")
    print(f"  Type: {mut['mutation_description']}")

Core Concepts

COSMIC Data Categories

Category	Description	Example
Cancer Gene Census	Curated list of cancer genes	TP53, KRAS, BRCA1
Somatic Mutations	Point mutations, indels	TP53 R248W
Copy Number	Amplifications, deletions	ERBB2 amplification
Gene Fusions	Translocation-derived fusions	BCR-ABL1
Mutational Signatures	Patterns of base changes	SBS1 (clock-like), SBS4 (smoking)
Drug Resistance	Mutations conferring resistance	EGFR T790M

Cancer Gene Census

Copy
# The Cancer Gene Census — curated cancer genes
def get_cancer_gene_census(headers):
    """Retrieve the COSMIC Cancer Gene Census"""
    url = "https://cancer.sanger.ac.uk/cosmic/api/v1/cancer-gene-census"
    response = requests.get(url, headers=headers)
    census = response.json()

    oncogenes = [g for g in census if "oncogene" in g.get("role", "").lower()]
    tsgs = [g for g in census if "tsg" in g.get("role", "").lower()]

    print(f"Total cancer genes: {len(census)}")
    print(f"Oncogenes: {len(oncogenes)}")
    print(f"Tumor suppressors: {len(tsgs)}")
    return census

# Lookup specific gene
def gene_mutation_profile(gene, headers):
    """Get mutation profile for a cancer gene"""
    url = f"https://cancer.sanger.ac.uk/cosmic/api/v1/gene/{gene}"
    response = requests.get(url, headers=headers)
    return response.json()

Mutational Signatures

Copy
# COSMIC Mutational Signatures (SBS, DBS, ID)
signatures = {
    "SBS1": "Spontaneous deamination (age-related, clock-like)",
    "SBS2": "APOBEC activity",
    "SBS4": "Tobacco smoking",
    "SBS6": "Defective DNA mismatch repair (MSI)",
    "SBS7a": "UV light exposure (melanoma)",
    "SBS10a": "POLE proofreading deficiency",
    "SBS13": "APOBEC activity (alternative)",
    "SBS22": "Aristolochic acid exposure",
}

# Dominant signatures by cancer type
cancer_signatures = {
    "melanoma": ["SBS7a", "SBS7b"],
    "lung_squamous": ["SBS4", "SBS2"],
    "colorectal_MSI": ["SBS6", "SBS15"],
    "breast": ["SBS1", "SBS2", "SBS13"],
}

Configuration

Parameter	Description	Default
api_token	COSMIC API authentication token	Required
genome_build	GRCh37 or GRCh38	GRCh38
result_limit	Max results per query	100
tumour_site	Filter by cancer type	None (all)
mutation_type	SNV, insertion, deletion, complex	None (all)

Best Practices

1. Use the Cancer Gene Census as your starting gene list. The CGC is expertly curated — start with known cancer genes rather than searching the full COSMIC database. It distinguishes oncogenes from tumor suppressors, guiding interpretation.

2. Filter by primary tissue type. A mutation's significance varies by cancer type. KRAS G12D is common in pancreatic cancer but rare in melanoma. Always contextualize mutation frequencies by the specific cancer type under investigation.

3. Check sample count, not just mutation presence. COSMIC reports how many samples carry each mutation. A mutation found in 1 sample out of 50,000 screened is very different from one found in 1,000. Use frequency data to prioritize significant mutations.

4. Combine COSMIC with functional databases. COSMIC catalogues mutations but doesn't always assess their functional impact. Cross-reference top mutations with functional data from DepMap (essentiality), OncoKB (clinical actionability), or ClinVar (pathogenicity).

5. Account for detection bias. Highly studied genes (TP53, KRAS) have more data than rarely studied genes. High mutation counts may reflect high screening frequency rather than biological importance. Consider the number of samples screened for each gene.

Common Issues

API access denied or rate limited. COSMIC requires registration for API access. Free academic accounts have rate limits — add delays between requests. For large-scale data needs, download the COSMIC data files directly rather than using the API.

Mutation nomenclature doesn't match between databases. COSMIC uses its own mutation IDs (COSV/COSM) alongside HGVS nomenclature. When cross-referencing with ClinVar or gnomAD, map via genomic coordinates (chromosome, position, ref, alt) rather than mutation names.

Different cancer type classifications between studies. COSMIC uses specific primary site and histology classifications that may not match other databases' terminology. Lung cancer in COSMIC splits into multiple subtypes — search by primary site first, then refine by histology.